The Tenth Circuit has proposed a rule under which anyone presenting a filing would certify that a human independently verified any generative AI assisted document before it reached the court

What proposed Rule 46.5(B)(4) would add

Rule 46.5(B) is the Tenth Circuit's existing representations-to-court clause. By presenting a brief, motion or other paper, whether by signing, filing, submitting or later advocating it, an attorney or unrepresented party already certifies three things, to the best of the person's knowledge, information and belief, formed after an inquiry reasonable under the circumstances.

The proposal, which is out for comment and binds nobody yet, would add a fourth. If adopted, the presenter would also certify that "any document prepared with the assistance of generative artificial intelligence" had been reviewed before filing or submission to the court "by a human who has independently verified the existence of all cited authorities; the accuracy of all quotations, citations, and legal analysis; and the document's compliance with all relevant rules and ethical obligations."

The redline shows the whole of item (4) as inserted text. It also moves the word "and" from the end of item (2) to the end of item (3), which is how the list grows from three items to four.

The Clerk's memo gives one sentence of purpose for the change: "The proposed revisions to subsections (B) and (C) of Tenth Circuit Rule 46.5 are intended to address the increased use of generative artificial intelligence by both lawyers and pro se litigants."

What the proposed text does not ask for

It is not a disclosure requirement. As drafted, item (4) would not require a filer to state that AI was used, name a tool, or attach a separate certificate. The promise would ride on the act of presenting the paper, as the existing three certifications do.

The draft does not define generative artificial intelligence. Across the 235 page redline, that phrase appears only in proposed item (4).

Two drafting points are ours, not the court's. First, the text asks that "a human" verified the document; it does not say the verifier has to be the signer. Second, item (4) would sit under the same knowledge-and-belief chapeau as the other three items, and how the court would read that qualifier against a duty to verify "all cited authorities" is untested because the rule is still a proposal.

The duty would also reach further than citation checking. Verification would cover legal analysis and compliance with "all relevant rules and ethical obligations", which is a wider promise than confirming that cases exist.

Sanctions: what the redline actually marks as new

We checked the rendered redline pages, not only the extracted text, because text extraction loses the strike and underline marking. On the page image, three pieces of proposed Rule 46.5(C) are shown as inserted: the words "but not limited to" after "including", item (1) "waiver of the affected issue", and item (2) "striking of a non-compliant document".

Dismissal or affirmance of the appeal is not new. Its wording appears as existing text, with only the new label (3) inserted. Monetary sanctions, initiation of disciplinary proceedings under the Plan for Attorney Disciplinary Enforcement, and an order to pay the other side's reasonable expenses including attorney's fees are also existing text, renumbered from (2), (3) and (4) to (4), (5) and (6).

If the proposal is adopted, subsection (C) would apply whenever a paper "is signed in violation of this rule". That covers all of Rule 46.5, not just the AI item, and the list of sanctions would become expressly non-exhaustive. The court would retain its discretion; nothing in the proposed text makes any sanction automatic.

A cross-reference the redline leaves at (C)(4)

Section 3.4 of the Plan for Attorney Disciplinary Enforcement, reproduced at the back of the redline, excludes from the Plan proceedings for damages, costs, expenses or attorney's fees under 28 U.S.C. 1927, Fed. R. App. P. 38, or Rule 46.5(C)(4). The redline edits the citation form on that line, changing "10th Cir. R." to "Rule", but leaves the number at (C)(4).

Under the proposed renumbering, (C)(4) would be monetary sanctions and the fee order would move to (C)(6). Whether that is intended or an oversight is not something the court has addressed. It is the kind of point a comment could raise, and we flag it as our reading of the redline, not as a stated problem.

Dates, and a conflict on the court's own pages

The signed memo and the court's news notice are both dated 18 September 2026, and they disagree on three details. The memo says comments may be submitted "through October 18, 2026"; the news page says "from October 18, 2026". The memo says the final "2027" rules will be posted "no later than November 30, 2026"; the news page refers to the final "2026" rules and a date of "no later than December 1, 2026".

We treat the signed memo as the operative statement, because it is the document the notice attaches, and the news page wording reads as an editing slip. The court has not corrected either page as far as we saw. Anyone planning to comment should send the comment before 18 October and not wait for that date.

Both documents say the changes "will take effect on January 1, 2027" while also inviting comment on them and describing them as proposed revisions. We read that as the planned start date for whatever is finally adopted, not as confirmation that this text will be adopted unchanged.

What we did not verify

What we opened: the full 235 page redline of the proposed 2027 rules, read as extracted text and, for Rule 46.5 and Plan section 3.4, as rendered page images at printed pages 152, 153 and 219. We also opened the two page Clerk's memo and the court's news notice of 18 September 2026, both fetched from the court's website.

What we did not open: any comments submitted, any committee report or statement of reasons beyond the memo's one sentence, and any general order or standing order of the Tenth Circuit on generative AI. We have not compared the proposal with the local rules of other circuits in this piece.

What we refuse to claim: we do not say the rule will be adopted, or adopted in this form, and we do not say it applies to any filing today. We do not describe dismissal, affirmance, monetary sanctions, disciplinary referral or fee orders as new, because the redline marks them as existing text. We do not say the Plan cross-reference is an error; we say it is unchanged. Quotations use ASCII apostrophes, and the em dashes the rule text places around "before filing or submission to the court" are paraphrased rather than reproduced, a house typesetting convention that changes no word.

Informational analysis for working professionals, not legal advice. Confirm how any rule applies to your situation with qualified counsel.

Related briefings

• Tenth Circuit publishes its AI citation rule, no sanctions
• Ninth Circuit AI sanctions ruling: what filings require
• AI court rules: sanctions, standing orders and filing duties
• North America AI regulation tracker
• AI Regulation News hub